arXiv subjects

Cheng Zhang

Publications and source records attributed to Cheng Zhang.

At least 91 records · Page 5Linked to original sources

arXiv2026

All Languages Matter: Understanding and Mitigating Language Bias in Multilingual RAG

Multilingual Retrieval-Augmented Generation (mRAG) leverages cross-lingual evidence to ground Large Language Models (LLMs) in global knowledge. However, we show that current mRAG systems suffer from a language bias during reranking, systematically favoring English and the query's native language. By introducing an estimated oracle evidence analysis, we quantify a substantial performance gap between existing rerankers and the achievable upper bound. Further analysis reveals a critical distributional mismatch: while optimal predictions require evidence scattered across multiple languages, current systems systematically suppress such ``answer-critical'' documents, thereby limiting downstream generation performance. To bridge this gap, we propose \textit{\textbf{L}anguage-\textbf{A}gnostic \textbf{U}tility-driven \textbf{R}eranker \textbf{A}lignment (LAURA)}, which aligns multilingual evidence ranking with downstream generative utility. Experiments across diverse languages and generation models show that LAURA effectively mitigates language bias and consistently improves mRAG performance.

cs.CL↗

arXiv2026

Near-Field Wideband Channel Estimation for XL-MIMO Systems via Denoising Diffusion Model

Extremely large-scale multiple-input multiple-output (XL-MIMO) is a key enabling technology for sixth-generation (6G) communication systems. Nevertheless, the increase in array aperture and signal bandwidth brings new challenges to wideband channel estimation in XL-MIMO systems. Motivated by recent advances in deep generative modeling, we propose a diffusion model-based method for near-field wideband channel estimation in XL-MIMO systems. We first analyze the statistical correlation of wideband channel and show that near-field wideband channel exhibits both spatial non-stationarity and beam split effects. Based on these observations, the channel estimation problem is formulated as a Bayesian posterior inference task, in which a diffusion model is employed to learn the prior distribution of the channel. To further enhance the representation of complex spatial-frequency channel structures, we design a denoising network with a multi-scale attention mechanism. In particular, the network extracts multi-scale spatial-frequency features via parallel convolutional branches with different receptive fields, and combines feature attention and spatial attention modules to adaptively emphasize critical channel features. This design enables more accurate modeling of near-field wideband channel distributions and consequently improves channel estimation performance. Experimental results demonstrate that the proposed method exhibits superior robustness to existing baseline schemes for XL-MIMO wideband channel estimation under different experimental settings.

eess.SP↗

arXiv2026

HeLa-Mem: Hebbian Learning and Associative Memory for LLM Agents

Long-term memory is a critical challenge for Large Language Model agents, as fixed context windows cannot preserve coherence across extended interactions. Existing memory systems represent conversation history as unstructured embedding vectors, retrieving information through semantic similarity. This paradigm fails to capture the associative structure of human memory, wherein related experiences progressively strengthen interconnections through repeated co-activation. Inspired by cognitive neuroscience, we identify three mechanisms central to biological memory: association, consolidation, and spreading activation, which remain largely absent in current research. To bridge this gap, we propose HeLa-Mem, a bio-inspired memory architecture that models memory as a dynamic graph with Hebbian learning dynamics. HeLa-Mem employs a dual-level organization: (1) an episodic memory graph that evolves through co-activation patterns, and (2) a semantic memory store populated via Hebbian Distillation, wherein a Reflective Agent identifies densely connected memory hubs and distills them into structured, reusable semantic knowledge. This dual-path design leverages both semantic similarity and learned associations, mirroring the episodic-semantic distinction in human cognition. Experiments on LoCoMo demonstrate superior performance across four question categories while using significantly fewer context tokens. Code is available on GitHub: https://github.com/ReinerBRO/HeLa-Mem

cs.CL↗

arXiv2026

Variational Autoencoding Discrete Diffusion with Enhanced Dimensional Correlations Modeling

Discrete diffusion models have recently shown great promise for modeling complex discrete data, with masked diffusion models (MDMs) offering a compelling trade-off between quality and generation speed. MDMs denoise by progressively unmasking multiple dimensions from an all-masked input, but their performance can degrade when using few denoising steps due to limited modeling of inter-dimensional dependencies. In this paper, we propose Variational Autoencoding Discrete Diffusion (VADD), a novel framework that enhances discrete diffusion with latent variable modeling to implicitly capture correlations among dimensions. By introducing an auxiliary recognition model, VADD enables stable training via variational lower bounds maximization and amortized inference over the training set. Our approach retains the efficiency of traditional MDMs while significantly improving sample quality, especially when the number of denoising steps is small. Empirical results on 2D toy data, pixel-level image generation, and text generation demonstrate that VADD consistently outperforms MDM baselines in sample quality with few denoising steps.

cs.LG↗

arXiv2026

DarwinTOD: LLM-driven Lifelong Self-evolution for Task-oriented Dialog Systems

Traditional task-oriented dialog systems are unable to evolve from ongoing interactions or adapt to new domains after deployment, that is a critical limitation in real-world dynamic environments. Continual learning approaches depend on episodic retraining with human curated data, failing to achieve autonomy lifelong improvement. While evolutionary computation and LLM driven self improvement offer promising mechanisms for dialog optimization, they lack a unified framework for holistic, iterative strategy refinement. To bridge this gap, we propose DarwinTOD, a lifelong self evolving dialog framework that systematically integrates these two paradigms, enabling continuous strategy optimization from a zero-shot base without task specific fine-tuning. DarwinTOD maintains an Evolvable Strategy Bank and operates through a dual-loop process: online multi-agent dialog execution with peer critique, and offline structured evolutionary operations that refine the strategy bank using accumulated feedback. This closed-loop design enables autonomous continuous improvement without human intervention. Extensive experiments show that DarwinTOD surpasses previous state-of-the-art methods and exhibits continuous performance gains throughout evolution. Our work provides a novel framework for building dialog systems with lifelong self evolution capabilities.

cs.MA↗

arXiv2026

Combating the Memory Walls: Optimization Pathways for Long-Context Agentic LLM Inference

LLMs now form the backbone of AI agents across a diverse range of applications, including tool use, command-line interfaces, and web or computer interaction. These agentic LLM inference tasks are fundamentally different from chatbot-focused inference. They often involve much longer context lengths to capture complex and prolonged inputs, such as an entire webpage DOM or complicated tool-call trajectories. This, in turn, generates significant off-chip memory traffic during inference and causes workloads to be constrained by two memory walls, namely the bandwidth wall and the capacity wall, preventing compute units from achieving high utilization. In this paper, we introduce PLENA, a hardware-software co-designed system built around three core optimization pathways. PLENA features a novel flattened systolic-array architecture (Pathway 1) and efficient compute and memory units that support an asymmetric quantization scheme (Pathway 2). It also provides native support for FlashAttention (Pathway 3). In addition, PLENA includes a complete software-hardware stack, consisting of a custom ISA, a compiler, a transaction-level simulator, and an automated design-space exploration flow. Experimental results show that PLENA delivers up to 2.23x and 4.70x higher throughput than the A100 GPU and TPU v6e, respectively, under identical multiplier counts and memory configurations during LLaMA agentic inference. PLENA also achieves up to 4.04x higher energy efficiency than the A100 GPU. The full PLENA system, including its simulator, compiler, ISA, and RTL implementation, will be open-sourced to the research community.

cs.AR↗

arXiv2026

A Decomposition Perspective to Long-context Reasoning for LLMs

Long-context reasoning is essential for complex real-world applications, yet remains a significant challenge for Large Language Models (LLMs). Despite the rapid evolution in long-context reasoning, current research often overlooks the internal complexity of the long-context reasoning task itself. In this paper, we move beyond this holistic view and decompose long-context reasoning into a set of fundamental atomic skills, and we then automatically synthesize a suite of pseudo datasets, each explicitly targeting a specific atomic skill. Our empirical analysis confirms that proficiency in these atomic skills is strongly correlated with general long-text reasoning performance. Building on this insight, we employ reinforcement learning on these pseudo datasets to sharpen the model's atomic skills, in the hope of boosting its general long-context reasoning ability. Extensive experiments across multiple benchmarks demonstrate the effectiveness of our approach: it outperforms a strong baseline by an average margin of 7.7\% (improving from 46.3\% to 54.0\%) across Loogle, Loong, LongBench-v2, BrowscompLong, Ruler-qa2, and MRCR.

cs.CL↗

arXiv2026

JoyAI-LLM Flash: Advancing Mid-Scale LLMs with Token Efficiency

We introduce JoyAI-LLM Flash, an efficient Mixture-of-Experts (MoE) language model designed to redefine the trade-off between strong performance and token efficiency in the sub-50B parameter regime. JoyAI-LLM Flash is pretrained on a massive corpus of 20 trillion tokens and further optimized through a rigorous post-training pipeline, including supervised fine-tuning (SFT), Direct Preference Optimization (DPO), and large-scale reinforcement learning (RL) across diverse environments. To improve token efficiency, JoyAI-LLM Flash strategically balances \emph{thinking} and \emph{non-thinking} cognitive modes and introduces FiberPO, a novel RL algorithm inspired by fibration theory that decomposes trust-region maintenance into global and local components, providing unified multi-scale stability control for LLM policy optimization. To enhance architectural sparsity, the model comprises 48B total parameters while activating only 2.7B parameters per forward pass, achieving a substantially higher sparsity ratio than contemporary industry leading models of comparable scale. To further improve inference throughput, we adopt a joint training-inference co-design that incorporates dense Multi-Token Prediction (MTP) and Quantization-Aware Training (QAT). We release the checkpoints for both JoyAI-LLM-48B-A3B Base and its post-trained variants on Hugging Face to support the open-source community.

cs.CL↗

arXiv2026

Shear and bulk viscosities of the gluon plasma across the transition temperature from lattice QCD

We investigate the temperature dependence of the shear viscosity ($η$) and bulk viscosity ($ζ$) of the gluon plasma using lattice QCD over the range 0.76--2.25$\,T_c$, extending from below the transition temperature $T_c$ across the transition region and into the deconfined phase. At each temperature, we employ three large, fine lattices, which enables controlled continuum extrapolations of the energy-momentum tensor correlators. Using gradient flow together with a recently developed blocking technique, we achieve percent-level precision for these correlators, providing strong constraints for a model-based spectral analysis. Since the inversion to real-time information is intrinsically ill posed, we extract viscosities by fitting spectral functions whose ultraviolet behavior is matched to the best available perturbative result, while the infrared region is described by a Lorentzian transport peak. The dominant modeling uncertainty associated with the transport peak width is bracketed by varying it over a physically motivated range set by thermal scales. We find that the shear-viscosity-to-entropy-density ratio, $η/s$, exhibits a minimum near the transition temperature $T_c$ and increases for $T>T_c$, whereas the bulk-viscosity-to-entropy-density ratio, $ζ/s$, decreases monotonically over the entire temperature range studied.

hep-lat↗

arXiv2026

A Nontrivial Upper Bound on the Out-of-Sample $R^2$ in Return Forecasting

This study establishes a nontrivial upper bound on the out-of-sample $R^2$ ($R^2_{\text{OOS}}$) in return forecasting. In particular, we define a coin-flip oracle model that, under the same directional accuracy, theoretically outperforms practical models in terms of MSE. The $R^2_{\text{OOS}}$ of the oracle model, whose analytical expression is a quadratic function of directional accuracy, can therefore serve as a tractable upper bound on the actual $R^2_{\text{OOS}}$. Empirical analyses across multiple forecasting scenarios reveal that the $R^2_{\text{OOS}}$ values of common predictive models are fundamentally bounded by this quadratic function.

econ.EM↗

arXiv2026

Towards Next-Generation Healthcare: A Survey of Medical Embodied AI for Perception, Decision-Making, and Action

Foundation models have demonstrated impressive performance in enhancing healthcare efficiency across a wide range of medical applications. Nevertheless, their limited ability to perceive, understand, and interact with the physical world significantly constrains their effectiveness in real-world clinical workflows, where safety-critical decision-making and physical execution are tightly coupled. Recently, embodied artificial intelligence (AI) has emerged as a promising physical-interactive paradigm for intelligent healthcare, enabling agents to operate in complex medical environments. As research in this area rapidly expands, understanding how intelligent agents function as integrated, end-to-end systems in clinical environments becomes increasingly critical. However, existing surveys on medical embodied AI largely emphasize individual aspects or functional components, lacking a unified system-level organization of the field. To support and consolidate recent advances, we systematically survey the core components of medical embodied AI, with a particular emphasis on the coordinated integration of perception, decision-making, and action. We further review representative medical applications and relevant datasets, and we analyze the major challenges encountered in real-world clinical practice. Finally, we discuss key directions for future research in this rapidly evolving field. The associated project can be found at https://github.com/VMVLab/Medical_Embodied_AI_Paper_List.

cs.AI↗

arXiv2026

Amortized Inference of Causal Models via Conditional Fixed-Point Iterations

Structural Causal Models (SCMs) offer a principled framework to reason about interventions and support out-of-distribution generalization, which are key goals in scientific discovery. However, the task of learning SCMs from observed data poses formidable challenges, and often requires training a separate model for each dataset. In this work, we propose an amortized inference framework that trains a single model to predict the causal mechanisms of SCMs conditioned on their observational data and causal graph. We first use a transformer-based architecture for amortized learning of dataset embeddings, and then extend the Fixed-Point Approach (FiP) to infer the causal mechanisms conditionally on their dataset embeddings. As a byproduct, our method can generate observational and interventional data from novel SCMs at inference time, without updating parameters. Empirical results show that our amortized procedure performs on par with baselines trained specifically for each dataset on both in and out-of-distribution problems, and also outperforms them in scarce data regimes.

cs.LG↗

arXiv2026

When the Server Steps In: Calibrated Updates for Fair Federated Learning

Federated learning (FL) has emerged as a transformative distributed learning paradigm, enabling multiple clients to collaboratively train a global model under the coordination of a central server without sharing their raw training data. While FL offers notable advantages, it faces critical challenges in ensuring fairness across diverse demographic groups. To address these fairness concerns, various fairness-aware debiasing methods have been proposed. However, many of these approaches either require modifications to clients' training protocols or lack flexibility in their aggregation strategies. In this work, we address these limitations by introducing EquFL, a novel server-side debiasing method designed to mitigate bias in FL systems. EquFL operates by allowing the server to generate a single calibrated update after receiving model updates from the clients. This calibrated update is then integrated with the aggregated client updates to produce an adjusted global model that reduces bias. Theoretically, we establish that EquFL converges to the optimal global model achieved by FedAvg and effectively reduces fairness loss over training rounds. Empirically, we demonstrate that EquFL significantly mitigates bias within the system, showcasing its practical effectiveness.

cs.LG↗

arXiv2026

Proactive Agent Research Environment: Simulating Active Users to Evaluate Proactive Assistants

Proactive agents that anticipate user needs and autonomously execute tasks hold great promise as digital assistants, yet the lack of realistic user simulation frameworks hinders their development. Existing approaches model apps as flat tool-calling APIs, failing to capture the stateful and sequential nature of user interaction in digital environments and making realistic user simulation infeasible. We introduce Proactive Agent Research Environment (Pare), a framework for building and evaluating proactive agents in digital environments. Pare models applications as finite state machines with stateful navigation and state-dependent action space for the user simulator, enabling active user simulation. Building on this foundation, we present Pare-Bench, a benchmark of 143 diverse tasks spanning communication, productivity, scheduling, and lifestyle apps, designed to test context observation, goal inference, intervention timing, and multi-app orchestration.

cs.AI↗

arXiv2026

TaxaAdapter: Vision Taxonomy Models are Key to Fine-grained Image Generation over the Tree of Life

Accurately generating images across the Tree of Life is difficult: there are over 10M distinct species on Earth, many of which differ only by subtle visual traits. Despite the remarkable progress in text-to-image synthesis, existing models often fail to capture the fine-grained visual cues that define species identity, even when their outputs appear photo-realistic. To this end, we propose TaxaAdapter, a simple and lightweight approach that incorporates Vision Taxonomy Models (VTMs) such as BioCLIP to guide fine-grained species generation. Our method injects VTM embeddings into a frozen text-to-image diffusion model, improving species-level fidelity while preserving flexible text control over attributes such as pose, style, and background. Extensive experiments demonstrate that TaxaAdapter consistently improves morphology fidelity and species-identity accuracy over strong baselines, with a cleaner architecture and training recipe. To better evaluate these improvements, we also introduce a multimodal Large Language Model-based metric that summarizes trait-level descriptions from generated and real images, providing a more interpretable measure of morphological consistency. Beyond this, we observe that TaxaAdapter exhibits strong generalization capabilities, enabling species synthesis in challenging regimes such as few-shot species with only a handful of training images and even species unseen during training. Overall, our results highlight that VTMs are a key ingredient for scalable, fine-grained species generation.

cs.CV↗

arXiv2026

Unified Camera Positional Encoding for Controlled Video Generation

Transformers have emerged as a universal backbone across 3D perception, video generation, and world models for autonomous driving and embodied AI, where understanding camera geometry is essential for grounding visual observations in three-dimensional space. However, existing camera encoding methods often rely on simplified pinhole assumptions, restricting generalization across the diverse intrinsics and lens distortions in real-world cameras. We introduce Relative Ray Encoding, a geometry-consistent representation that unifies complete camera information, including 6-DoF poses, intrinsics, and lens distortions. To evaluate its capability under diverse controllability demands, we adopt camera-controlled text-to-video generation as a testbed task. Within this setting, we further identify pitch and roll as two components effective for Absolute Orientation Encoding, enabling full control over the initial camera orientation. Together, these designs form UCPE (Unified Camera Positional Encoding), which integrates into a pretrained video Diffusion Transformer through a lightweight spatial attention adapter, adding less than 1% trainable parameters while achieving state-of-the-art camera controllability and visual fidelity. To facilitate systematic training and evaluation, we construct a large video dataset covering a wide range of camera motions and lens types. Extensive experiments validate the effectiveness of UCPE in camera-controllable video generation and highlight its potential as a general camera representation for Transformers across future multi-view, video, and 3D tasks. Code will be available at https://github.com/chengzhag/UCPE.

cs.CV↗

arXiv2026

MiniBEE: A New Form Factor for Compact Bimanual Dexterity

Bimanual robot manipulators can achieve impressive dexterity, but typically rely on two full six- or seven- degree-of-freedom arms so that paired grippers can coordinate effectively. This traditional framework increases system complexity while only exploiting a fraction of the overall workspace for dexterous interaction. We introduce the MiniBEE (Miniature Bimanual End-effector), a compact system in which two reduced-mobility arms (3+ DOF each) are coupled into a kinematic chain that preserves full relative positioning between grippers. To guide our design, we formulate a kinematic dexterity metric that enlarges the dexterous workspace while keeping the mechanism lightweight and wearable. The resulting system supports two complementary modes: (i) wearable kinesthetic data collection with self-tracked gripper poses, and (ii) deployment on a standard robot arm, extending dexterity across its entire workspace. We present kinematic analysis and design optimization methods for maximizing dexterous range, and demonstrate an end-to-end pipeline in which wearable demonstrations train imitation learning policies that perform robust, real-world bimanual manipulation.

cs.RO↗

arXiv2026

Probing How Scalable Table Data Enhances General Long-Context Reasoning

As real-world tasks grow increasingly complex, long-context reasoning has become a core capability for Large Language Models (LLMs). However, few studies explore which data types are effective for long-context reasoning and why. We find that structured table data with periodic structures shows strong potential for long-context reasoning. Motivated by this observation, we mathematically analyze tabular dependency structures using mutual information, revealing periodic non-vanishing dependencies in table data. Furthermore, we systematically analyze the capabilities of structured table data, conduct relevant scaling experiments, and validate its underlying mechanisms for enhancing long-context reasoning, yielding several meaningful insights. Leveraging these insights, we propose a simple yet scalable pipeline(TableLong) for synthesizing high-quality, diverse, and verifiable structured table data to boost long-context reasoning via RL. Extensive experimental results demonstrate that table data significantly enhances the long-context reasoning capability of LLMs across multiple long-context benchmarks (+8.24\% on average), and even improves performance on out-of-domain benchmarks (+8.06\% on average). We hope that our insights provide practical guidance for effective post-training data to enhance long-context reasoning in LLMs.

cs.CL↗